GitHub Adds GPT-6 Sol and Luna to Copilot, Giving Coders More Model Choices

GitHub has added OpenAI’s GPT-6 Sol and GPT-6 Luna to Copilot, expanding its GPT-6 lineup beyond Astra. Sol is aimed at interactive, multistep coding, while Luna handles smaller, faster jobs at the family’s lowest cost.

The two models join GPT-6 Astra, which GitHub made generally available in Copilot earlier this month. GitHub positions Astra for long-horizon autonomous coding and agentic tasks. Sol and Luna add two more choices, letting users match a model to the work instead of treating GPT-6 as a single general-purpose setting.

A middle option and a budget option

GitHub calls Sol a balanced model for interactive and agentic coding. Its intended role is development work that benefits from careful, multistep validation—work where an assistant needs to move through several checks rather than return a quick answer.

Luna serves a different purpose. GitHub describes it as lightweight and cost-efficient for smaller, faster tasks, making it the lowest-cost option in the GPT-6 family. The announcement does not say which coding jobs should move to Luna, so developers will still need to decide when speed and cost matter more than a fuller workflow.

The choice arrives across Copilot’s main surfaces

Eligible users can select Sol and Luna in Visual Studio Code, Visual Studio, Copilot CLI, GitHub’s Copilot cloud agent, the Copilot app, GitHub Mobile, JetBrains, Xcode and Eclipse. That places the same model decision in an editor, command-line workflow, mobile app and cloud-based agent environment rather than confining it to one product surface.

Access differs by plan and policy

Luna is available to Copilot Pro, Pro+, Max, Business and Enterprise plans. Sol is available to Pro+, Max, Business and Enterprise plans, leaving standard Pro subscribers with access to Luna but not Sol. Both models use usage-based billing, so choosing between the models can affect more than the kind of answer a developer receives.

For organizations, the model picker is not an entirely individual choice. Business and Enterprise administrators can manage GPT-6 access through Copilot’s model policies. Under GitHub’s default enablement setting, new models are enabled automatically unless an administrator has disabled the global default or the particular model.

Availability is not immediate for everyone

GitHub says the rollout is gradual, meaning an eligible account may not see either option immediately. The update expands Copilot’s menu, but it does not resolve the harder practical question: when is a coding task modest enough for a lower-cost model, and when does it warrant a model built for more deliberate validation?